An Evaluation of Risk Metrics

Vanguard Investment Counseling & Research

Executive summary Author

Frank J. Ambrosio, CFA
Risk assessment is a critical element of effective investing.

Risks can be measured in many different ways.

Appropriate risk-measurement techniques depend on specific portfolio objectives.

Even the best risk measures have limitations. Awareness of these limitations can
help investors avoid common pitfalls.

Risk is the uncertainty of an outcome. Risk metrics quantify this uncertainty, providing
an estimate of the range of possible outcomes. These measures are critical tools for
portfolio construction and performance assessment because a principal assumption
of investing is that to achieve a given level of return, investments with lower risk
are preferred over those with higher risk. Investments that carry more risk typically
offer a higher level of expected returns to compensate for the greater uncertainty.
However, increased risk by its very nature means that these higher expected
returns may not be realized in a given time period. Like any other tool, risk metrics
have limitations.

C o n n e c t w i t h Va n g u a rd > www.vanguard.com
 Figure 1. Range of calendar-year returns for various stock Figure 2. Various absolute risk measures for investment returns
and bond allocations, 19262006
Historical or simulated
frequency of returns (%)
60% Risk of loss
(probability of Mean
negative returns) return
40
Average annual return (%)

20 Shortfall risk
Annual return (%)

7.8 8.3 8.9 9.3 9.7 10.1 10.5 Bell-curve distribution

5.2 5.9 6.6 7.2 (probability of return

(a shape of likely outcomes)
below minimum
0 target rate)

20
Value-at-risk
(99% VaR)
40
Asset
60 returns
Return at 1% 0% Target Mean
Stocks 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% probability return return
Bonds 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
Asset allocation Standard deviation (a measure of dispersion around the mean;
one standard deviation encompasses about 68% of returns)
Benchmarks
Stocks = Standard & Poors 500 Index, 19261970; Dow Jones Wilshire 5000 Source: Vanguard Investment Counseling & Research.
Composite Index, January 1, 1971April 22, 2005; Morgan Stanley Capital
International (MSCI) US Broad Market Index, April 23, 2005December 31, 2006.
Bonds = Ibbotson Intermediate Government series, 19261972; Lehman Brothers
Intermediate U.S. Treasury Index, 19732006.

Sources: Ibbotson Associates, Vanguard Investment Counseling & Research.

Risk metrics are based on the historical performance
of asset classes. Although historical performance is
inescapably the foundation for our best estimates of
future risk, the future will never repeat the past. The
proverbial one-hundred-year storm can undermine
expectations based on even the most sophisticated
risk metrics.
Even if risk cant be predicted with certainty, risk
metrics provide critical information to help answer
the most important question for any investor: How
should a portfolio be invested to gain the best
chance of achieving its objectives? Risk metrics have
also proven reliable for comparing the relative risks
of different asset classes. For example, Figure 1
displays the range of annual returns for various
combinations of stocks and bonds over 81 years,
providing a clear illustration that as the weighting of
stocks in a portfolio increases, so does the volatility
of the portfolios returns.
Risk measures fall into two categories: absolute
and relative. In this paper, we review widely used
metrics, describing their uses and limitations. We
also identify considerations for investors applying
risk metrics to portfolio construction and oversight.
Ideally, for major mandates such as pension or
endowment funds, risk metrics will be incorporated
into the policy statement. This practice not only
provides guidance to portfolio managers but can also
help fiduciaries in assessing portfolio performance
over time.
Successful use of risk metrics depends on selecting
measures that are consistent with a portfolios
objectives. It is also essential to understand the
limitations of available data.

2 > Vanguard Investment Counseling & Research

Absolute risk measures
Widely used measures of absolute risk include
standard deviation, value-at-risk (VaR), risk of loss,
and shortfall risk. Here we define these measures
and illustrate their uses and limitations.
Portfolio standard deviation
Standard deviation, a basic statistical tool, is widely
used to measure the degree of fluctuation in a
portfolios return. The larger the standard deviation,
the greater the magnitude of the fluctuations from
the portfolios average return. Consider a portfolio
with an average return of 10% and a standard
deviation of 15%. Its returns should fall between
5% and 25% in 68% of all observations.
Standard deviation can be a useful measure for
portfolios such as pension and endowment funds
that are concerned with the consequences of both
positive and negative deviations from a specific
return target or portfolio value. It is less appropriate
for investors who worry more about negative returns
than about upside deviations. In addition, standard
deviation assumes that returns are normally
distributed, which limits its use for investments
with unusual return distributions.
Historical value-at-risk
Historical value-at-risk is based on a securitys
worst results over a given period. The measure can
be derived from a fixed percentage of the worst
observationsthe worst 1% or 5%, for example
or a fixed number of those observations.
Historical VaR can provide a visceral sense of what
risk means. The worst annual return for the stock
market in the last 81 years was 43.1% (in 1931). In
other words, the stock markets one-year VaR, based
on roughly the worst 1% of observations, is 43.1%.
This metric can be useful in explaining risk to people
who are less familiar with investment theory.
Because historical VaR looks only at the worst
observations without any regard for their frequency,
many analysts prefer to use it in combination with
other measures. Historical VaR is often employed
in combination with risk of loss.
Risk of loss
Risk of loss is a useful counterpart to historical VaR,
describing the frequency of bad results. It measures
the percentage of outcomes below a certain total-
return level, usually 0%. This risk metric is most
often used to assess the likelihood that a portfolio
will fall below a specific return or asset value target.
Value-at-risk and risk of loss can be used by almost
any portfolio as the ultimate test of risk tolerance.
Would there be comfort if an extreme negative
event occurred, or should risk levels be reduced?
Suppose an investors objective is to maintain a
portfolios nominal value for a year. A risk-of-loss
metric indicating, for example, that during the past
81 years, 30% of annual stock market returns were
negative would be a useful piece of information.
Vanguard Investment Counseling & Research > 3
 Table 1. Example of risk metrics using broad asset-class data for 19262006

Nominal total return Real total return

Percentage of Percentage of
Average years with Highest Average years with Highest
annual negative one-year annual negative one-year
Asset class return return loss return return loss
Treasury bills 3.8% 0% 0.0% 0.8% 35% 15.0%
Bonds (intermediate-term Treasury) 5.2% 9% 2.3% 2.1% 38% 14.5%
Stocks 10.5% 30% 43.1% 7.2% 35% 37.3%
Benchmarks
Treasury bills: Citigroup 3-Month Treasury Bill Index, 19262006.
Bonds: Ibbotson Intermediate Government series, 19261972; Lehman Intermediate U.S. Treasury Index, 19732006.
Stocks: S&P 500 Index, 19261970; Dow Jones Wilshire 5000 Index, January 1, 1971April 22, 2005; MSCI US Broad Market Index, April 23, 2005December 31, 2006.
Sources: Ibbotson Associates, Vanguard Investment Counseling & Research.

Risk of loss can also be measured in real terms: Shortfall risk

that is, the amount of the investments return
after inflation is removed. Table 1 shows that the
measurement of losses in real terms can be quite
different from the measurement in nominal terms.
Historical VaR and risk of loss are most useful
when they are derived from results over long time
periods. If based on shorter, less-volatile periods,
these measures may give investors a false sense
of security. Generally, the longer the period, the
more negative the worst potential loss could be at
any pointsomething that is true looking either
backward or forward in time.
Shortfall risk is the probability that an investments
value will be less than is needed to meet portfolio
objectives. This probability can be measured using
a variety of approaches, including historical time-
path simulation and Monte Carlo simulation. The
shortfall-risk metric is most often used to help
create a comprehensive investment plan, taking
into account current assets and estimated future
liabilities in both the accumulation and spending
phases. The metric might indicate, for example,
that an investment portfolio stands a 25% chance
of being depleted before the end of the liability
funding period.

4 > Vanguard Investment Counseling & Research

Shortfall risk could be used by an institution or
individual spending, or expecting to spend, regularly
from portfolio assets. Examples include endow-
ments and foundations, pension plans, and people
investing for retirement objectives.
Although the result is a seemingly simple probability
percentage, the measure can be extremely compli-
cated to calculate and understand. Minor changes to
inputs and assumptions can cause wide differences
in the results. Shortfall risk is only as good as the
inputs, most of which are uncertain and subject to
estimation errors.
Relative risk measures
Widely used measures of relative risk include excess
return, tracking error, Sharpe ratio, information ratio,
beta, and Treynor ratio.
Excess return
Excess return is a securitys return above or below
that of a benchmark or a theoretically risk-free asset
such as Treasury bills. Excess return is simply the
portfolios or securitys return minus that of the
benchmark, making it easy to understand and
calculate. If an actively managed mutual fund returns
11%, and its benchmark returns 10%, the funds
excess return is 1%.
Investment advisors use excess return to compare
their portfolio returns with those of a benchmark.
The metrics usefulness depends on two assump-
tions: that the portfolios total risk (systematic and
idiosyncratic) is similar to the benchmarks, and that
the returns of the portfolio and benchmark always
move in the same direction. Unless these conditions
are met, excess return will provide limited insight.
A portfolio might outperform a benchmark, for
example, but take on more risk to do so. Excess
return wouldnt account for this risk.
Tracking error
Tracking error is the standard deviation of excess
return. Like portfolio standard deviation, tracking
error assumes that returns are normally distributed.
It combines both upside and downside risk. Consider
an index fund that has no excess return relative to
its benchmark when measured over a long period,
but that produces an annualized tracking error of
10 basis points (0.1 percentage point). If the bench-
mark returns 10% per year, the funds return should
be between 9.9% and 10.1% (10 basis points on
either side of the benchmark return of 10%) in
68% of the observed one-year time periods (one
standard deviation).
Tracking error is commonly used to assess an index
funds success in matching its target index. Active
managers who are closely tied to a benchmark
(quantitative managers, for example) might describe
their expected deviation from the benchmark in
terms of tracking error. The metric could also be
used to evaluate the success of a risk-controlled
mandate, such as a balanced portfolio designed to
generate returns within 50 basis points of a balanced
benchmarks performance. Like portfolio standard
deviation, tracking error is less useful to investors
concerned mostly with downside risk.
Tracking error also is less directly relevant for
actively managed portfolios that are not closely
pegged to benchmarks. However, this metric is
used in calculating the information ratio (discussed
on the next page), which is often employed in
comparisons of active managers.
Vanguard Investment Counseling & Research > 5
 Sharpe ratios are used to compare investments
Table 2. Sharpe ratios of various assets over portions from the same asset class or from asset classes
of the period 19702006 with similar liquidity and valuation characteristics.
Sharpe ratios are extremely time-period dependent
Sharpe ratios
and can vary greatly depending upon the length of
19701981 19821999 20002006
the period used, as shown in Table 2.
Commodities 0.3984 0.1654 0.3356
Real estate 0.2577 0.4742 1.3920
In addition, an uncritical use of Sharpe ratios can
International developed
stock markets 0.0841 0.4809 0.0911
lead to faulty conclusions. As Table 2 indicates,
higher ratios often occur when an assets per-
U.S. stock market 0.0301 0.7381 0.0806
formance is reaching its peakas was true for
U.S. fixed income market 0.1375 0.8193 0.9032
commodities in 1981 and the stock market in 1999.
Long U.S. Treasury bonds 0.2743 0.5647 0.5914
If the Sharpe ratio were being used to select an
Benchmarks
Commodities: Goldman Sachs Commodity Index (GSCI) Total Return Index.
asset for future outperformance, the metric could
Real estate: National Association of Real Estate Investment Trusts (NAREIT) Equity REIT Index. be counterproductive.
International stock markets: MSCI Europe, Australasia, Far East (EAFE) Index.
U.S. stock market: S&P 500 Index, 1970; Dow Jones Wilshire 5000 Index, January 1, 1971
April 22, 2005; MSCI US Broad Market Index, April 23, 2005December 31, 2006. Information ratio
U.S. fixed income market: Citigroup High Grade Bond Index, 19701975; Lehman Aggregate The information ratio is the risk-adjusted return of a
Bond Index, 19762006.
Long U.S. Treasury bonds: Ibbotson Long Government series, 19701972; Lehman Long portfolio or security versus a benchmark. To calculate
U.S. Treasury Index, 19732006. the information ratio, an assets excess return is
Source: Vanguard Investment Counseling & Research. divided by its tracking error relative to the benchmark.
(The Sharpe ratio is actually an information ratio that
uses the risk-free return as the benchmark.) This
metric is typically used to measure a managers skill
Sharpe ratio versus peers.
The Sharpe ratio is a representation of the risk-
adjusted return of a portfolio or security. The An actively managed fund that has 100 basis points
Sharpe ratio measures how much return is being of excess return and 200 basis points of tracking
obtained for each theoretical unit of risk. To calculate error relative to its benchmark would have an
a Sharpe ratio, an assets excess return versus a information ratio of 0.5. A fund that had the same
risk-free asset such as Treasury bills is divided by excess return and 400 basis points of tracking error
the standard deviation of the assets returns. would have an information ratio of 0.25. All else
equal, higher information ratio values are preferred.
Sharpe ratios can be negative if the asset under-
performs the risk-free asset. For a longer-term As with the Sharpe ratio, information ratios are
measurement, the ratio generally falls in a range extremely time-period dependent and can vary
from 0 to +1; the higher the number, the better. greatly depending upon the length of the period
used. They tend to peak at the times of best
performance, sending potentially misleading
signals to investors about future returns.

6 > Vanguard Investment Counseling & Research

Beta An actively managed equity fund with a 1% excess
Beta is the magnitude of an investments price return and a beta of 1.20 would have a Treynor
fluctuations relative to the ups and downs of the ratio of 0.833. Higher Treynor ratios are preferred
overall market. The market (or index) is assigned to lower ones.
a beta of 1.00; if a portfolio has a beta of 1.20, its
price would rise or fall by 12% when the market The Treynor ratio is based on beta, so it shares betas
rises or falls by 10%. Beta is best used to measure limitations. In addition, the Treynor ratio magnifies
the systematic or market risk of an investment and small differences in beta over time, particularly in the
can be appropriate in evaluating an investment for case of low-beta strategies such as those of market-
possible inclusion in a diversified portfolio. neutral hedge funds. Dramatic changes in Treynor
ratios may not necessarily reflect dramatic changes
Hedge fund managers often cite beta as a in risk.
measure of portfolio risk. For example, market-
neutral managers try to offset the betas of their Pitfalls of risk measures
long positions with those of their short positions in a portfolio context
to reduce the portfolios overall beta toward zero.
Risk metrics are critical tools for portfolio
In calculating betas, correct benchmark selection
construction and oversight, but a simplistic use
is critical. A portfolio beta derived from an index
of these measures can lead to faulty conclusions.
with different risk characteristics reveals little about
Potential pitfalls are related to the limitations of
the portfolios potential volatility. Another caution:
historical data, the integrity of the data, and
Betas are not stable over time. Frequent, even
challenges in adapting metrics to a specific portfolio.
daily, rebalancing may be necessary to maintain
Maintaining awareness of what risk metrics can
a portfolios beta.
tell youand what they cantis the best way to
sidestep these pitfalls.
Treynor ratio
The Treynor ratio is the risk-adjusted return of
Limitations of history
a portfolio or security versus the market. It is
All risk metrics use historical data to make
an assets excess return versus a risk-free asset
assumptions about future risk. But what if the
such as Treasury bills, divided by the assets beta.
future proves to be very different from the past?
The Treynor ratio is an appropriate measure of a
Until October 19, 1987, history suggested that the
portfolios return per unit of risk. The measure has
historical VaR for any one-day period in the U.S.
a drawback, however, in that it assumes a portfolio
stock market was 12.8%the decline of the Dow
manager has diversified away all of the unsystematic
Jones Industrial Average (DJIA) on October 28, 1929.
(company-specific) risk, and that only systematic
On October 19, 1987, the DJIA fell by 22.6%, and
(market) risk is left. This limits the use of the
the Wilshire 5000 Index lost 17.2% of its value.
Treynor ratio to comparisons involving extremely
well-diversified portfolios.

Vanguard Investment Counseling & Research > 7

 Did stock market risk increase dramatically
overnight? For an investor concerned with the
worst possible one-day outcome, the answer might
be yes. However, most portfolio managers take a
longer-term view: Many mandates have a horizon
measured in decades. For these portfolios, risk
did not change quite so dramatically, especially
considering that the total return of equities for all
of 1987 was positive.
Data integrity
Data can present an incomplete picture of the
opportunities available to real-world investors. In
less-liquid markets, such as emerging markets or
the small-cap stock arena, for example, higher
transaction and management costs diminish the
opportunities implied by cost-free benchmarks and
risk metrics. This difference between expectations
and reality can lead to flawed conclusions about an
assets optimal weight or a managers success.
Quantitative risk metrics may also imply a precision
and reliability that dont exist in the underlying data.
Many investment-strategy databases have survivor
bias and incubation bias that can overstate the
performance of different strategies and managers.
These risks are magnified when performance
reporting is voluntary. In addition, the returns of
some strategiesprivate equity, hedge fundsare
based on relatively infrequent appraisal or mark-to-
market methods, which can keep volatility artificially
low. A study co-authored by Burton Malkiel shows
that the full universe of hedge-fund returns, if
made public and computed, would be 400500
basis points lower than the reported returns
(Malkiel, 2005).
8 > Vanguard Investment Counseling & Research
Time-period dependency
As the October 1987 crash suggests, risk is
time-period dependent. Risk metrics based on
a longer time series are less influenced by short-
term extremes, but the entire history of our
financial markets is, in essence, just one time
period. How do we address the risk of time-
period dependency?
First, we could forecast risk through modeling
based on a combination of historical and current
information. This approach has its own short-
comings, however, depending upon the length of
the forecast period and the number and accuracy
of the variables involved. While the use of current
data could possibly enhance accuracy, these short-
comings introduce an additional forecasting risk
that must be taken into account with any solution.
Second, we could simulate the future using the
available historical data by employing either a time-
pathing or a Monte Carlo approach. The time-pathing
approach keeps the historical data series in the order
in which it occurred, but each simulation starts at a
different point in time. As the series reaches the end
date, the simulations loop around to the beginning of
the data to fill in the end of the investment horizon.
This approach is useful for selecting a subset from
a long series of historical data.
The Monte Carlo approach uses many more
simulations and fits the simulated data to the
characteristics of the historical data, but does not
use the actual historical sequences themselves.
This approach can highlight the potential for extreme
outcomes that did not actually occur in history. The
Monte Carlo approach is useful for historical data
that is distributed fairly normally, generally in a bell-
shaped curve. This approach can employ additional
techniques, such as vector auto-regression, to help
ensure that the results include not only the normalized
distribution of the returns, but also a sense of the
environment in which these returns are achieved.
Ultimately, selecting the appropriate time period for
a portfolios risk metrics can become a question not
only of time horizon but also of how the financial
markets performance is modeled over that period.
Benchmark selection and manager risk
Benchmark selection is an equally tricky adaptation
issue. Relative risk measures such as excess return,
tracking error, and beta are based in part on the
performance of a benchmark. If the benchmark
does not represent the portfolio effectively, these
risk measures will fail to provide useful information.
Consider, for example, an equity mutual fund that is
being compared with the Standard & Poors 500
Index. The S&P 500 is dominated by large-
capitalization stocks. If the equity fund happens to
invest only in mid-cap stocks, its characteristics will
differ significantly from those of the benchmark. If
the fund further devotes itself to growth stocks,
there will be an additional style difference between
the benchmark and the fund. These differences
would show up as a larger tracking error versus the
benchmark and a more volatile excess return. Beta,
which would have a value of 1 if the benchmark
exactly matched the fund, might deviate significantly
from 1 and would vary with time, possibly switching
above and below 1.
Someone evaluating this mid-cap growth fund on
the basis of the S&P 500 comparison might
conclude that active management was responsible
for the high degree of variation. Note, however, that
the benchmark mismatch could produce similar
results for a mid-cap growth index fund.
The question of benchmark selection is directly
connected to the process of portfolio design. Ideally,
portfolios are constructed in a top-down process
that is, the investor first clarifies the investment
objective and establishes the level of risk that is
deemed acceptable. The investor then determines
the mix of asset classes most suitable to the
objective, next chooses sub-asset classes, and finally
selects specific investment vehicles. Among the
many benefits of this process is that the investor
gains a deep understanding of the portfolio and thus
can readily evaluate which benchmarks are relevant,
and which less so, for its performance.
Investors who lack such understanding of their
portfolios are vulnerable to being misled by metrics
that do not account for risks absent from market
benchmarks, such as manager risk, factor risk, and
investment-specific risk. In particular, choosing
investment vehicles that are heavily dependent on
a managers performance may expose the flaws of
portfolio-analysis tools. The manager selected for
a strategy is a major influence on returns when the
performance spread between managers is wide,
as in the private equity, hedge fund, or real estate
markets. In that situation, using index returns as
a benchmark can lead to unrealistic expectations
and to portfolio allocations that are unproductive
over time.
By the same token, using quantitative methods on
their own can lead investors into some serious traps.
A degree of qualitative adjustment to the inputs
is appropriate and prudent when comparing asset
classes over several different lengths of time, when
juxtaposing different valuation measurements,
and, most importantly, when choosing vehicles to
implement specific investment strategies.
Vanguard Investment Counseling & Research > 9
 Precise objective is key to success
The more precisely you define your objective, the
better your ability to select metrics that can help
you evaluate your portfolios performance and risk
exposures. Examples of objectives and possible
primary risk measurements include:
Objective: Reach a targeted real-return level
to meet current annual distributions. Possible
key risk measure: Find the investment or
portfolio with the lowest percentage of real-
return observations below the targeta
risk-of-loss metric.
Objective: Minimize present value of future
investment contributions. Possible key risk
measure: Measure standard deviation of
returns relative to a target return.
Objective: Accumulate wealth to meet a
future level of real liabilities. Possible key risk
measure: Measure the shortfall risk of each
potential investment or portfolio and choose
the one with the lowest probability of missing
the target liability level.
Objective: Match the return of a specified
benchmark. Possible key risk measure: Measure
the tracking error of the portfolio versus that
benchmark.
Objective: Maximize risk-adjusted performance.
Possible key risk measure: Look for an investment
or combination of investments with the highest
Sharpe ratio.
10 > Vanguard Investment Counseling & Research
Conclusion
Risk measurement is a vital part of portfolio
construction and oversight. The investment
policy statement should articulate a portfolios
risk-management goals and identify the metrics
used to evaluate its success in reaching them.
The appropriate metrics depend on the portfolios
objective. In addition to choosing the right
quantitative tools, it is important to use them
with a qualitative understanding of their limitations.
Blind reliance on sometimes-flawed measures can
lead to suboptimal investment decisions.
The difficulty in measuring risk only serves to
underscore the value of a top-down approach to
asset allocation. Because major asset classes like
stocks, bonds, and cash have long return histories
and well-defined relative risk relationships, portfolio
construction should begin with a decision about
how best to combine these classes based on the
portfolios objective and the level of risk deemed
acceptable. Decisions about specific investments
should be made only at the end of this process and
again should incorporate a risk analysis. Portfolio
benchmarks should be carefully chosen to match the
investment strategy. Following the process of first
determining the objectives, then determining the
risk measures, will result in a deeper understanding
of the risks being taken in the investment portfolio
and a better ability to assess its subsequent
performance.
Additional reading/references
Clarke, Roger G., 1998. Alternative Measures of Risk.
In Investment Management, Peter L. Bernstein and
Aswath Damodaran (eds.). New York: John Wiley &
Sons. p. 81-98.
Hirt, Geoffrey A., and Stanley B. Block, 2006.
Fundamentals of Investment Management. Boston,
Mass.: McGraw-Hill/Irwin. 710 p.
Malkiel, Burton G., and Atanu Saha, 2005. Hedge
Funds: Risk and Return. Financial Analysts Journal
61(6):80-8.
Nawrocki, David N., 1999. A Brief History of
Downside Risk Measures. Journal of Investing
8(3):9-25.
Sharpe, William F., 1994. The Sharpe Ratio.
Journal of Portfolio Management 21(1):49-58.
Stocks, Mary Ellen, and Christopher to, 1997.
Value at Risk for the Asset Manager. Journal of
Performance Measurement 1(4):35-43.

Molitor, Jeffrey S., 2006. Evaluating managers:

Are we sending the right messages? Valley
Forge, Pa.: Investment Counseling & Research,
The Vanguard Group. 13 p.

Vanguard Investment Counseling & Research > 11

 P.O. Box 2600
Valley Forge, PA 19482-2600

C o n n e c t w i t h Va n g u a rd > www.vanguard.com

Connect with Vanguard, Vanguard, and the ship logo are Vanguard Investment Counseling & Research
trademarks of The Vanguard Group, Inc. All other marks are
the exclusive property of their respective owners. Ellen Rinaldi, J.D., LL.M./Principal/Department Head
Joseph H. Davis, Ph.D./Principal
Francis M. Kinniry Jr., CFA/Principal
Frank J. Ambrosio, CFA
John Ameriks, Ph.D.
Donald G. Bennyhoff, CFA
Maria Bruno, CFP
Scott J. Donaldson, CFA, CFP
Michael Hess
Julian Jackson
Colleen M. Jaconetti, CFP, CPA
Kushal Kshirsagar, Ph.D.
Karin Peterson LaBarge, Ph.D.
Christopher B. Philips, CFA
Glenn Sheay, CFA
Kimberly A. Stockton
Yesim Tokat, Ph.D.
David J. Walker, CFA

2007 The Vanguard Group, Inc.

All rights reserved.

FLGERM 0307